HS project

Zoom in hair follicle stem cells (HFSCs)

This file is used to generate a dataset containing only hair follicle stem cells (HFSCs).

library(dplyr)
library(patchwork)
library(ggplot2)

.libPaths()

## [1] "/usr/local/lib/R/library"

Preparation

In this section, we set the global settings of the analysis. We will store data there :

save_name = "hfsc"
out_dir = "."

We load the sample information :

sample_info = readRDS(paste0(out_dir, "/../../1_metadata/hs_hd_sample_info.rds"))
project_names_oi = sample_info$project_name

graphics::pie(rep(1, nrow(sample_info)),
              col = sample_info$color,
              labels = sample_info$project_name)

Here are custom colors for each cell type :

color_markers = readRDS(paste0(out_dir, "/../../1_metadata/hs_hd_color_markers.rds"))

data.frame(cell_type = names(color_markers),
           color = unlist(color_markers)) %>%
  ggplot2::ggplot(., aes(x = cell_type, y = 0, fill = cell_type)) +
  ggplot2::geom_point(pch = 21, size = 5) +
  ggplot2::scale_fill_manual(values = unlist(color_markers), breaks = names(color_markers)) +
  ggplot2::theme_classic() +
  ggplot2::theme(legend.position = "none",
                 axis.line = element_blank(),
                 axis.title = element_blank(),
                 axis.ticks = element_blank(),
                 axis.text.y = element_blank(),
                 axis.text.x = element_text(angle = 30, hjust = 1))

Make hfsc dataset

Atlas

We load the combined dataset containing all cell types from all samples :

sobj = readRDS(paste0(out_dir, "/../../3_combined/hs_hd_sobj.rds"))
sobj

## An object of class Seurat 
## 20003 features across 12111 samples within 1 assay 
## Active assay: RNA (20003 features, 2000 variable features)
##  6 dimensional reductions calculated: RNA_pca, RNA_pca_38_tsne, RNA_pca_38_umap, harmony, harmony_38_umap, harmony_38_tsne

We represent cells in the tSNE :

name2D = "harmony_38_tsne"

We smooth cell type annotation at a cluster level :

cluster_type = table(sobj$cell_type, sobj$seurat_clusters) %>%
  prop.table(., margin = 2) %>%
  apply(., 2, which.max)
cluster_type = setNames(nm = names(cluster_type),
                        levels(sobj$cell_type)[cluster_type])

sobj$cluster_type = cluster_type[sobj$seurat_clusters]

We look gene markers expression level, cell annotation and cluster-smoothed annotation on the projection, to locate hfsc cells :

hfsc_markers = c("KRT15", "DIO2", "TCEAL2")
hfsc_cell_type = c("HFSC")
color_markers[!(names(color_markers) %in% hfsc_cell_type)] = "gray92"

# Feature Plot
plot_list = lapply(hfsc_markers, FUN = function(one_gene) {
  p = Seurat::FeaturePlot(sobj, reduction = name2D,
                          features = one_gene) +
    Seurat::NoAxes() +
    ggplot2::scale_color_gradientn(colors = aquarius:::color_gene) +
    ggplot2::theme(aspect.ratio = 1,
                   plot.subtitle = element_text(hjust = 0.5))
  return(p)
})

# Cell type annotation
plot_list[[length(plot_list) + 1]] = Seurat::DimPlot(sobj, group.by = "cell_type",
                                                     cols = color_markers, reduction = name2D,
                                                     order = save_name) +
  ggplot2::labs(title = "Cell annotation",
                subtitle = paste0(sum(sobj$cell_type %in% hfsc_cell_type),
                                  " cells")) +
  Seurat::NoAxes() + Seurat::NoLegend() +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5),
                 plot.subtitle = element_text(hjust = 0.5))

# Cluster-smoothed annotation
plot_list[[length(plot_list) + 1]] = Seurat::DimPlot(sobj,
                                                     reduction = name2D,
                                                     group.by = "cluster_type") +
  ggplot2::scale_color_manual(values = c(unname(unlist(color_markers[hfsc_cell_type])),
                                         rep("gray92", length(color_markers) - length(hfsc_cell_type))),
                              breaks = c(hfsc_cell_type, setdiff(names(color_markers), hfsc_cell_type))) +
  ggplot2::labs(title = "Cluster annotation",
                subtitle = paste0(sum(sobj$cluster_type %in% hfsc_cell_type),
                                  " cells")) +
  Seurat::NoAxes() + Seurat::NoLegend() +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5),
                 plot.subtitle = element_text(hjust = 0.5))

patchwork::wrap_plots(plot_list, nrow = 1)

Combined dataset

We extract cells of interest based on the clustering :

sobj = subset(sobj, cluster_type %in% hfsc_cell_type)
sobj

## An object of class Seurat 
## 20003 features across 1454 samples within 1 assay 
## Active assay: RNA (20003 features, 2000 variable features)
##  6 dimensional reductions calculated: RNA_pca, RNA_pca_38_tsne, RNA_pca_38_umap, harmony, harmony_38_umap, harmony_38_tsne

We remove all things that were calculated based on the full atlas :

sobj = Seurat::DietSeurat(sobj)
sobj

## An object of class Seurat 
## 20003 features across 1454 samples within 1 assay 
## Active assay: RNA (20003 features, 2000 variable features)

Clean metadata

We keep a subset of meta.data and reset levels :

sobj@meta.data = sobj@meta.data[, c("orig.ident", "nCount_RNA", "nFeature_RNA", "log_nCount_RNA",
                                    "project_name", "sample_identifier", "sample_type", "cell_type",
                                    "Seurat.Phase", "cyclone.Phase", "percent.mt", "percent.rb")]

sobj$orig.ident = factor(sobj$orig.ident, levels = levels(sample_info$project_name))
sobj$project_name = factor(sobj$project_name, levels = levels(sample_info$project_name))
sobj$sample_identifier = factor(sobj$sample_identifier, levels = levels(sample_info$sample_identifier))
sobj$sample_type = factor(sobj$sample_type, levels = levels(sample_info$sample_type))

summary(sobj@meta.data)

##    orig.ident    nCount_RNA     nFeature_RNA  log_nCount_RNA    project_name
##  2021_31: 29   Min.   :  866   Min.   : 502   Min.   : 6.764   2021_31: 29  
##  2021_36: 19   1st Qu.: 5674   1st Qu.:2001   1st Qu.: 8.644   2021_36: 19  
##  2021_41:435   Median : 9290   Median :2737   Median : 9.137   2021_41:435  
##  2022_03:395   Mean   :10544   Mean   :2751   Mean   : 9.044   2022_03:395  
##  2022_14:334   3rd Qu.:14492   3rd Qu.:3513   3rd Qu.: 9.581   2022_14:334  
##  2022_01: 32   Max.   :47543   Max.   :6746   Max.   :10.770   2022_01: 32  
##  2022_02:210                                                   2022_02:210  
##  sample_identifier sample_type         cell_type    Seurat.Phase      
##  HS_1: 29          HS:1212     HFSC         :1286   Length:1454       
##  HS_2: 19          HD: 242     ORS          : 104   Class :character  
##  HS_3:435                      IBL          :  41   Mode  :character  
##  HS_4:395                      proliferative:   9                     
##  HS_5:334                      cuticle      :   4                     
##  HD_1: 32                      IFE          :   3                     
##  HD_2:210                      (Other)      :   7                     
##  cyclone.Phase        percent.mt       percent.rb     
##  Length:1454        Min.   : 0.000   Min.   : 0.4948  
##  Class :character   1st Qu.: 3.387   1st Qu.:19.7097  
##  Mode  :character   Median : 4.562   Median :24.1822  
##                     Mean   : 5.839   Mean   :22.5877  
##                     3rd Qu.: 6.586   3rd Qu.:28.4049  
##                     Max.   :19.826   Max.   :43.0197  
## 

Processing

Metadata

How many cells by sample ?

table(sobj$project_name)

## 
## 2021_31 2021_36 2021_41 2022_03 2022_14 2022_01 2022_02 
##      29      19     435     395     334      32     210

We represent this information as a piechart :

graphics::pie(table(sobj$project_name),
              col = sample_info$color,
              labels = sample_info$project_name)

Pprojection

We remove genes that are expressed in less than 5 cells :

sobj = aquarius::filter_features(sobj, min_cells = 5)

## [1] 20003  1454
## [1] 15384  1454

sobj

## An object of class Seurat 
## 15384 features across 1454 samples within 1 assay 
## Active assay: RNA (15384 features, 1379 variable features)

We normalize the count matrix for remaining cells :

sobj = Seurat::NormalizeData(sobj,
                             normalization.method = "LogNormalize")
sobj = Seurat::FindVariableFeatures(sobj, nfeatures = 2000)
sobj = Seurat::ScaleData(sobj)

sobj

## An object of class Seurat 
## 15384 features across 1454 samples within 1 assay 
## Active assay: RNA (15384 features, 2000 variable features)

We perform a PCA :

sobj = Seurat::RunPCA(sobj,
                      assay = "RNA",
                      reduction.name = "RNA_pca",
                      npcs = 100,
                      seed.use = 1337L)
sobj

## An object of class Seurat 
## 15384 features across 1454 samples within 1 assay 
## Active assay: RNA (15384 features, 2000 variable features)
##  1 dimensional reduction calculated: RNA_pca

We choose the number of dimensions such that they summarize 35 % of the variability :

stdev = sobj@reductions[["RNA_pca"]]@stdev
stdev_prop = cumsum(stdev)/sum(stdev)
ndims = which(stdev_prop > 0.35)[1]
ndims

## [1] 24

We can visualize this on the elbow plot :

elbow_p = Seurat::ElbowPlot(sobj, ndims = 100, reduction = "RNA_pca") +
  ggplot2::geom_point(x = ndims, y = stdev[ndims], col = "red")
x_text = ggplot_build(elbow_p)$layout$panel_params[[1]]$x$get_labels() %>% as.numeric()
elbow_p = elbow_p +
  ggplot2::scale_x_continuous(breaks = sort(c(x_text, ndims)), limits = c(0, 100))
x_color = ifelse(ggplot_build(elbow_p)$layout$panel_params[[1]]$x$get_labels() %>%
                   as.numeric() %>% round(., 2) == round(ndims, 2), "red", "black")
elbow_p = elbow_p +
  ggplot2::theme_classic() +
  ggplot2::theme(axis.text.x = element_text(color = x_color))

elbow_p

Without correction

We generate a tSNE and a UMAP with 24 principal components :

sobj = Seurat::RunTSNE(sobj,
                       reduction = "RNA_pca",
                       dims = 1:ndims,
                       seed.use = 1337L,
                       reduction.name = paste0("RNA_pca_", ndims, "_tsne"))

sobj = Seurat::RunUMAP(sobj,
                       reduction = "RNA_pca",
                       dims = 1:ndims,
                       seed.use = 1337L,
                       reduction.name = paste0("RNA_pca_", ndims, "_umap"))

We can visualize the two representations :

tsne = Seurat::DimPlot(sobj, group.by = "project_name",
                       reduction = paste0("RNA_pca_", ndims, "_tsne")) +
  ggplot2::scale_color_manual(values = sample_info$color,
                              breaks = sample_info$project_name) +
  Seurat::NoAxes() + ggplot2::ggtitle("PCA - tSNE") +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5),
                 legend.position = "none")

umap = Seurat::DimPlot(sobj, group.by = "project_name",
                       reduction = paste0("RNA_pca_", ndims, "_umap")) +
  ggplot2::scale_color_manual(values = sample_info$color,
                              breaks = sample_info$project_name) +
  Seurat::NoAxes() + ggplot2::ggtitle("PCA - UMAP") +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5))

tsne | umap

There is a strong batch-effect.

Harmony

We remove batch-effect using Harmony :

`%||%` = function(lhs, rhs) {
  if (!is.null(x = lhs)) {
    return(lhs)
  } else {
    return(rhs)
  }
}

set.seed(1337L)
sobj = harmony::RunHarmony(object = sobj,
                           group.by.vars = "project_name",
                           plot_convergence = TRUE,
                           reduction = "RNA_pca",
                           assay.use = "RNA",
                           reduction.save = "harmony",
                           max.iter.harmony = 20,
                           project.dim = FALSE)

From this batch-effect removed projection, we generate a tSNE and a UMAP.

sobj = Seurat::RunUMAP(sobj, 
                       seed.use = 1337L,
                       dims = 1:ndims,
                       reduction = "harmony",
                       reduction.name = paste0("harmony_", ndims, "_umap"),
                       reduction.key = paste0("harmony_", ndims, "umap_"))
sobj = Seurat::RunTSNE(sobj,
                       dims = 1:ndims,
                       seed.use = 1337L,
                       reduction = "harmony",
                       reduction.name = paste0("harmony_", ndims, "_tsne"),
                       reduction.key = paste0("harmony", ndims, "tsne_"))

These are the corrected UMAP and tSNE :

tsne = Seurat::DimPlot(sobj, group.by = "project_name",
                       reduction = paste0("harmony_", ndims, "_tsne")) +
  ggplot2::scale_color_manual(values = sample_info$color,
                              breaks = sample_info$project_name) +
  Seurat::NoAxes() + ggplot2::ggtitle("PCA - harmony - tSNE") +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5),
                 legend.position = "none")

umap = Seurat::DimPlot(sobj, group.by = "project_name",
                       reduction = paste0("harmony_", ndims, "_umap")) +
  ggplot2::scale_color_manual(values = sample_info$color,
                              breaks = sample_info$project_name) +
  Seurat::NoAxes() + ggplot2::ggtitle("PCA - harmony - UMAP") +
  ggplot2::theme(aspect.ratio = 1,
                 plot.title = element_text(hjust = 0.5))

tsne | umap

We will keep the tSNE from Harmony :

reduction = "harmony"
name2D = paste0("harmony_", ndims, "_tsne")

Clustering

We generate a clustering :

sobj = Seurat::FindNeighbors(sobj, reduction = reduction, dims = 1:ndims)
sobj = Seurat::FindClusters(sobj, resolution = 0.5)

## Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
## 
## Number of nodes: 1454
## Number of edges: 56839
## 
## Running Louvain algorithm...
## Maximum modularity in 10 random starts: 0.8368
## Number of communities: 9
## Elapsed time: 0 seconds

dimplot_clusters = Seurat::DimPlot(sobj, reduction = name2D, label = TRUE) +
  Seurat::NoAxes() +
  ggplot2::theme(aspect.ratio = 1)
dimplot_clusters

Visualization

We can represent the 4 quality metrics :

plot_list = Seurat::FeaturePlot(sobj, reduction = name2D,
                                combine = FALSE, pt.size = 0.5,
                                features = c("percent.mt", "percent.rb", "nFeature_RNA", "log_nCount_RNA"))
plot_list = lapply(plot_list, FUN = function(one_plot) {
  one_plot +
    Seurat::NoAxes() +
    ggplot2::scale_color_gradientn(colors = aquarius:::color_gene) +
    ggplot2::theme(aspect.ratio = 1)
})

patchwork::wrap_plots(plot_list, nrow = 1)

Project name

We can visualize the two batch-effect corrected representations :

plot_list = lapply(c(paste0("harmony_", ndims, "_tsne"),
                     paste0("harmony_", ndims, "_umap")), FUN = function(one_proj) {
                       Seurat::DimPlot(sobj, group.by = "project_name",
                                       reduction = one_proj) +
                         ggplot2::scale_color_manual(values = sample_info$color,
                                                     breaks = sample_info$project_name) +
                         Seurat::NoAxes() + ggplot2::ggtitle(one_proj) +
                         ggplot2::theme(aspect.ratio = 1,
                                        plot.title = element_text(hjust = 0.5),
                                        legend.position = "none")
                     })

patchwork::wrap_plots(plot_list, ncol = 2)

Clusters

We can represent clusters, split by sample of origin :

plot_list = aquarius::plot_split_dimred(sobj,
                                        reduction = name2D,
                                        split_by = "sample_identifier",
                                        group_by = "seurat_clusters",
                                        split_color = setNames(sample_info$color,
                                                               nm = sample_info$sample_identifier),
                                        group_color = aquarius::gg_color_hue(length(levels(sobj$seurat_clusters))),
                                        main_pt_size = 0.5, bg_pt_size = 0.5)

plot_list[[length(plot_list) + 1]] = dimplot_clusters

patchwork::wrap_plots(plot_list, ncol = 4) +
  patchwork::plot_layout(guides = "collect") &
  ggplot2::theme(legend.position = "none")

We make a heatmap to see clusters distribution among samples :

cluster_markers = c("KRT15", "DIO2", "TCEAL2", "LGR5",
                    "ANGPTL7", "EPCAM", "KRT75", "COL7A1",
                    "PTHLH", "AQP3", "BDNF", "TGFB2",
                    # QC metrics
                    "percent.mt", "percent.rb", "log_nCount_RNA")

ht_annot = Seurat::FetchData(sobj, slot = "data", vars = cluster_markers) %>%
  as.data.frame()
ht_annot$clusters = sobj$seurat_clusters
ht_annot = ht_annot %>%
  dplyr::group_by(clusters) %>%
  dplyr::summarise_all(funs('mean' = mean)) %>%
  as.data.frame() %>%
  dplyr::select(-clusters) %>%
  `colnames<-`(c(cluster_markers))
head(ht_annot)

##      KRT15      DIO2    TCEAL2      LGR5    ANGPTL7     EPCAM      KRT75
## 1 3.772408 1.7594414 1.0967168 0.1531288 0.04813027 0.1488576 0.09599649
## 2 2.872929 0.9717549 0.4865434 0.8227170 0.02848058 0.9794145 0.08906393
## 3 3.266954 0.2330210 0.5469870 0.9311975 1.57926410 0.2195408 0.09673598
## 4 3.173199 1.5098691 0.3740698 0.6364927 0.15928182 0.1235805 0.10010731
## 5 3.955449 1.0164850 0.9865835 0.3527980 0.04785611 0.1588205 0.08017276
## 6 2.756580 0.5849426 0.2112691 0.9236196 0.02685297 0.5758125 2.22797150
##      COL7A1      PTHLH       AQP3       BDNF     TGFB2 percent.mt percent.rb
## 1 0.6193851 0.08013840 0.52304424 0.27311714 1.2476834   5.198222   24.15048
## 2 0.3291017 0.00626933 0.08261837 0.03737522 0.1808539   4.237108   28.88093
## 3 0.2800722 0.01687239 0.04917580 0.01747412 0.3483168   3.345700   24.01082
## 4 1.4946769 0.03439439 0.15593885 0.14610995 0.9322713  14.144596    4.39859
## 5 0.4448983 0.03153255 0.34839389 0.13812667 0.5417283   4.914900   27.47871
## 6 0.4427799 0.02054159 0.37649768 0.04783346 0.1188677   4.002230   23.81659
##   log_nCount_RNA
## 1       9.179242
## 2       9.431060
## 3       8.970504
## 4       8.052919
## 5       8.760704
## 6       9.522789

color_fun = function(one_gene) {
  gene_range = range(ht_annot[, one_gene])
  gene_palette = circlize::colorRamp2(colors = c("#FFFFFF", aquarius::color_gene[-1]),
                                      breaks = seq(from = gene_range[1], to = gene_range[2],
                                                   length.out = length(aquarius::color_gene)))
  return(gene_palette)
}

ha = ComplexHeatmap::HeatmapAnnotation(df = ht_annot,
                                       which = "column",
                                       show_legend = TRUE,
                                       col = setNames(nm = cluster_markers,
                                                      lapply(cluster_markers, FUN = color_fun)),
                                       annotation_name_side = "left")

ht = aquarius::plot_prop_heatmap(df = sobj@meta.data[, c("sample_identifier", "seurat_clusters")],
                                 bottom_annotation = ha,
                                 cluster_rows = TRUE,
                                 prop_margin = 1,
                                 row_names_gp = grid::gpar(names = sample_info$sample_identifier,
                                                           col = sample_info$color,
                                                           fontface = "bold"),
                                 row_title = "Sample",
                                 column_title = "Cluster")

ComplexHeatmap::draw(ht,
                     merge_legends = TRUE)

We also look at genes of interest on the projection :

plot_list = lapply(cluster_markers, FUN = function(one_gene) {
  p = Seurat::FeaturePlot(sobj, features = one_gene,
                          pt.size = 0.2, reduction = name2D) +
    ggplot2::scale_color_gradientn(colors = aquarius::color_gene) +
    Seurat::NoAxes() +
    ggplot2::theme(aspect.ratio = 1)
  
  return(p)
})

patchwork::wrap_plots(plot_list, ncol = 5)

Save

We save the Seurat object :

saveRDS(sobj, file = paste0(out_dir, "/", save_name, "_sobj.rds"))

R Session

show

## R version 3.6.3 (2020-02-29)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.6 LTS
## 
## Matrix products: default
## BLAS:   /usr/local/lib/R/lib/libRblas.so
## LAPACK: /usr/local/lib/R/lib/libRlapack.so
## 
## locale:
## [1] C
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] ggplot2_3.3.5   patchwork_1.1.2 dplyr_1.0.7    
## 
## loaded via a namespace (and not attached):
##   [1] softImpute_1.4              graphlayouts_0.7.0         
##   [3] pbapply_1.4-2               lattice_0.20-41            
##   [5] haven_2.3.1                 vctrs_0.3.8                
##   [7] usethis_2.0.1               dynwrap_1.2.1              
##   [9] blob_1.2.1                  survival_3.2-13            
##  [11] prodlim_2019.11.13          dynutils_1.0.5             
##  [13] later_1.3.0                 DBI_1.1.1                  
##  [15] R.utils_2.11.0              SingleCellExperiment_1.8.0 
##  [17] rappdirs_0.3.3              uwot_0.1.8                 
##  [19] dqrng_0.2.1                 jpeg_0.1-8.1               
##  [21] zlibbioc_1.32.0             pspline_1.0-18             
##  [23] pcaMethods_1.78.0           mvtnorm_1.1-1              
##  [25] htmlwidgets_1.5.4           GlobalOptions_0.1.2        
##  [27] future_1.22.1               UpSetR_1.4.0               
##  [29] laeken_0.5.2                leiden_0.3.3               
##  [31] clustree_0.4.3              parallel_3.6.3             
##  [33] scater_1.14.6               irlba_2.3.3                
##  [35] DEoptimR_1.0-9              tidygraph_1.1.2            
##  [37] Rcpp_1.0.9                  readr_2.0.2                
##  [39] KernSmooth_2.23-17          carrier_0.1.0              
##  [41] promises_1.1.0              gdata_2.18.0               
##  [43] DelayedArray_0.12.3         limma_3.42.2               
##  [45] graph_1.64.0                RcppParallel_5.1.4         
##  [47] Hmisc_4.4-0                 fs_1.5.2                   
##  [49] RSpectra_0.16-0             fastmatch_1.1-0            
##  [51] ranger_0.12.1               digest_0.6.25              
##  [53] png_0.1-7                   sctransform_0.2.1          
##  [55] cowplot_1.0.0               DOSE_3.12.0                
##  [57] here_1.0.1                  TInGa_0.0.0.9000           
##  [59] ggraph_2.0.3                pkgconfig_2.0.3            
##  [61] GO.db_3.10.0                DelayedMatrixStats_1.8.0   
##  [63] gower_0.2.1                 ggbeeswarm_0.6.0           
##  [65] iterators_1.0.12            DropletUtils_1.6.1         
##  [67] reticulate_1.26             clusterProfiler_3.14.3     
##  [69] SummarizedExperiment_1.16.1 circlize_0.4.15            
##  [71] beeswarm_0.4.0              GetoptLong_1.0.5           
##  [73] xfun_0.35                   bslib_0.3.1                
##  [75] zoo_1.8-10                  tidyselect_1.1.0           
##  [77] reshape2_1.4.4              purrr_0.3.4                
##  [79] ica_1.0-2                   pcaPP_1.9-73               
##  [81] viridisLite_0.3.0           rtracklayer_1.46.0         
##  [83] rlang_1.0.2                 hexbin_1.28.1              
##  [85] jquerylib_0.1.4             dyneval_0.9.9              
##  [87] glue_1.4.2                  RColorBrewer_1.1-2         
##  [89] matrixStats_0.56.0          stringr_1.4.0              
##  [91] lava_1.6.7                  europepmc_0.3              
##  [93] DESeq2_1.26.0               recipes_0.1.17             
##  [95] labeling_0.3                harmony_0.1.0              
##  [97] httpuv_1.5.2                class_7.3-17               
##  [99] BiocNeighbors_1.4.2         DO.db_2.9                  
## [101] annotate_1.64.0             jsonlite_1.7.2             
## [103] XVector_0.26.0              bit_4.0.4                  
## [105] mime_0.9                    aquarius_0.1.5             
## [107] Rsamtools_2.2.3             gridExtra_2.3              
## [109] gplots_3.0.3                stringi_1.4.6              
## [111] processx_3.5.2              gsl_2.1-6                  
## [113] bitops_1.0-6                cli_3.0.1                  
## [115] batchelor_1.2.4             RSQLite_2.2.0              
## [117] randomForest_4.6-14         tidyr_1.1.4                
## [119] data.table_1.14.2           rstudioapi_0.13            
## [121] org.Mm.eg.db_3.10.0         GenomicAlignments_1.22.1   
## [123] nlme_3.1-147                qvalue_2.18.0              
## [125] scran_1.14.6                locfit_1.5-9.4             
## [127] scDblFinder_1.1.8           listenv_0.8.0              
## [129] ggthemes_4.2.4              gridGraphics_0.5-0         
## [131] R.oo_1.24.0                 dbplyr_1.4.4               
## [133] BiocGenerics_0.32.0         TTR_0.24.2                 
## [135] readxl_1.3.1                lifecycle_1.0.1            
## [137] timeDate_3043.102           ggpattern_0.3.1            
## [139] munsell_0.5.0               cellranger_1.1.0           
## [141] R.methodsS3_1.8.1           proxyC_0.1.5               
## [143] visNetwork_2.0.9            caTools_1.18.0             
## [145] codetools_0.2-16            Biobase_2.46.0             
## [147] GenomeInfoDb_1.22.1         vipor_0.4.5                
## [149] lmtest_0.9-38               msigdbr_7.5.1              
## [151] htmlTable_1.13.3            triebeard_0.3.0            
## [153] lsei_1.2-0                  xtable_1.8-4               
## [155] ROCR_1.0-7                  BiocManager_1.30.10        
## [157] scatterplot3d_0.3-41        abind_1.4-5                
## [159] farver_2.0.3                parallelly_1.28.1          
## [161] RANN_2.6.1                  askpass_1.1                
## [163] GenomicRanges_1.38.0        RcppAnnoy_0.0.16           
## [165] tibble_3.1.5                ggdendro_0.1-20            
## [167] cluster_2.1.0               future.apply_1.5.0         
## [169] Seurat_3.1.5                dendextend_1.15.1          
## [171] Matrix_1.3-2                ellipsis_0.3.2             
## [173] prettyunits_1.1.1           lubridate_1.7.9            
## [175] ggridges_0.5.2              igraph_1.2.5               
## [177] RcppEigen_0.3.3.7.0         fgsea_1.12.0               
## [179] remotes_2.4.2               scBFA_1.0.0                
## [181] destiny_3.0.1               VIM_6.1.1                  
## [183] testthat_3.1.0              htmltools_0.5.2            
## [185] BiocFileCache_1.10.2        yaml_2.2.1                 
## [187] utf8_1.1.4                  plotly_4.9.2.1             
## [189] XML_3.99-0.3                ModelMetrics_1.2.2.2       
## [191] e1071_1.7-3                 foreign_0.8-76             
## [193] withr_2.5.0                 fitdistrplus_1.0-14        
## [195] BiocParallel_1.20.1         xgboost_1.4.1.1            
## [197] bit64_4.0.5                 foreach_1.5.0              
## [199] robustbase_0.93-9           Biostrings_2.54.0          
## [201] GOSemSim_2.13.1             rsvd_1.0.3                 
## [203] memoise_2.0.0               evaluate_0.18              
## [205] forcats_0.5.0               rio_0.5.16                 
## [207] geneplotter_1.64.0          tzdb_0.1.2                 
## [209] caret_6.0-86                ps_1.6.0                   
## [211] DiagrammeR_1.0.6.1          curl_4.3                   
## [213] fdrtool_1.2.15              fansi_0.4.1                
## [215] highr_0.8                   urltools_1.7.3             
## [217] xts_0.12.1                  GSEABase_1.48.0            
## [219] acepack_1.4.1               edgeR_3.28.1               
## [221] checkmate_2.0.0             scds_1.2.0                 
## [223] cachem_1.0.6                npsurv_0.4-0               
## [225] babelgene_22.3              rjson_0.2.20               
## [227] openxlsx_4.1.5              ggrepel_0.9.1              
## [229] clue_0.3-60                 rprojroot_2.0.2            
## [231] stabledist_0.7-1            tools_3.6.3                
## [233] sass_0.4.0                  nichenetr_1.1.1            
## [235] magrittr_2.0.1              RCurl_1.98-1.2             
## [237] proxy_0.4-24                car_3.0-11                 
## [239] ape_5.3                     ggplotify_0.0.5            
## [241] xml2_1.3.2                  httr_1.4.2                 
## [243] assertthat_0.2.1            rmarkdown_2.18             
## [245] boot_1.3-25                 globals_0.14.0             
## [247] R6_2.4.1                    Rhdf5lib_1.8.0             
## [249] nnet_7.3-14                 RcppHNSW_0.2.0             
## [251] progress_1.2.2              genefilter_1.68.0          
## [253] statmod_1.4.34              gtools_3.8.2               
## [255] shape_1.4.6                 HDF5Array_1.14.4           
## [257] BiocSingular_1.2.2          rhdf5_2.30.1               
## [259] splines_3.6.3               AUCell_1.8.0               
## [261] carData_3.0-4               colorspace_1.4-1           
## [263] generics_0.1.0              stats4_3.6.3               
## [265] base64enc_0.1-3             dynfeature_1.0.0           
## [267] smoother_1.1                gridtext_0.1.1             
## [269] pillar_1.6.3                tweenr_1.0.1               
## [271] sp_1.4-1                    ggplot.multistats_1.0.0    
## [273] rvcheck_0.1.8               GenomeInfoDbData_1.2.2     
## [275] plyr_1.8.6                  gtable_0.3.0               
## [277] zip_2.2.0                   knitr_1.41                 
## [279] ComplexHeatmap_2.14.0       latticeExtra_0.6-29        
## [281] biomaRt_2.42.1              IRanges_2.20.2             
## [283] fastmap_1.1.0               ADGofTest_0.3              
## [285] copula_1.0-0                doParallel_1.0.15          
## [287] AnnotationDbi_1.48.0        vcd_1.4-8                  
## [289] babelwhale_1.0.1            openssl_1.4.1              
## [291] scales_1.1.1                backports_1.2.1            
## [293] S4Vectors_0.24.4            ipred_0.9-12               
## [295] enrichplot_1.6.1            hms_1.1.1                  
## [297] ggforce_0.3.1               Rtsne_0.15                 
## [299] shiny_1.7.1                 numDeriv_2016.8-1.1        
## [301] polyclip_1.10-0             grid_3.6.3                 
## [303] lazyeval_0.2.2              Formula_1.2-3              
## [305] tsne_0.1-3                  crayon_1.3.4               
## [307] MASS_7.3-54                 pROC_1.16.2                
## [309] viridis_0.5.1               dynparam_1.0.0             
## [311] rpart_4.1-15                zinbwave_1.8.0             
## [313] compiler_3.6.3              ggtext_0.1.0